Find the bias your decisions don't show you.
Hiring, promotions, loan approvals, complaint handling, tenant screening: if a system or a scoring tool helps decide what happens to people, we check whether it treats some groups worse than others. One-off or every quarter, at a fixed price, and without you sending us anyone's personal data.
Any decision about people, made or helped by a system.
You don't need an AI team or a new platform. If you can export the decisions, we can review them.
Who moves forward
Résumé screeners, assessments, AI interviews and ranking tools. Including the figures NYC Local Law 144 asks for.
Who gets ahead
Promotions, performance ratings, bonuses and terminations, for the people you already employ.
Who gets approved
Loan approvals, credit limits and pricing, for fintechs, credit unions and small lenders.
Who gets heard
Which complaints are escalated, which claims are paid, and how long each group waits.
Who gets the keys
Tenant screening scores and application decisions, for property managers and screening providers.
"We don't collect race." Your model might not need it.
A system can learn race and ethnicity from data that looks neutral. Leaving the column out does not take the bias out.
Proxies hide in ordinary data
- Fields that look neutral can stand in for race or ethnicity
- A model trained on past decisions repeats their bias
- Nobody sees it, because the race column was never there
The gap, even without a race column
- Whether some groups are approved or selected less, and by how much
- Group-level results only, ready to share with counsel or a regulator
- No individual is labeled, and the estimate is never used to decide about anyone
In the example above, every applicant group had the same credit scores and incomes. The model only penalized a field that looked neutral. With no race column at all, the check shows Black applicants approved at 0.78 of the rate of White applicants.
Once, or every quarter. No software license.
Two prices. Nothing else to add.
Several systems or departments? Ask us for a quote.
A single review
$1,950 / system or department
When you need an answer now: before a launch, a renewal, an audit or a board meeting.
- Selection or approval rates and impact ratios by group
- Proxy bias check if you don't collect race or ethnicity
- For hiring tools, the Local Law 144 tables and what your public summary still needs, ready for your independent auditor
- Audit-ready documentation: a methodology note, a draft public summary and a draft candidate notice for your counsel to review
- PDF report with findings and recommendations, plus CSV, and a one-page summary for leadership
- A free re-check within 60 days, after you make changes
A review every quarter
$5,800 / year, per system or department
When decisions keep changing: new models, new policies, new volumes.
- Four full reviews a year, one each quarter, with your latest data and everything in the one-off
- Documentation kept up to date every quarter, ready for an audit or a complaint
- What changed since the last one: what improved, what drifted
- An early warning before a small gap becomes a pattern
- One payment for the year
Four steps to your report.
Buy the review
One-off or quarterly, paid before we start.
Share aggregates
Send us totals per group, or run Fairspire on your own computer and send us its results.
We review
Rates, ratios, proxy checks and the reasons behind the numbers.
You decide
A report your leadership, legal team and auditor can read, with what to fix first.
Start with one decision you'd like to be sure about.
Tell us what the system decides. We reply with the next steps.
- Sales sales@blackodix.com
- General hello@blackodix.com
- Security security@blackodix.com
Fairspire